2013-06-06 07:27 PM
I have a FAS3240 with 2 x shelves of 600GB SAS disks. I wish to place both VMware, NFS and EXCH 2010, FC on these shelves and just going through all my different aggr, RG and Vol options.
Disks Total 48
Disks Useable 38 = 20TB
FAS3240 Max RG Size 26D+2D = 14TB
To migrate from previous storage vendor to Netapp my VMware and Exchange data requirements before dedup are 6TB each.
I'm sure there is a better way to get the maximum available and performance on the storage so I was hoping someone could help verify the following configuration for me please. New to Netapp.
Vol_VMWare - 7.6TB
Vol_Exch - 8.2TB
Solved! SEE THE SOLUTION
2013-06-06 11:31 PM
Welcome to a Whole New World with NetApp!
First and foremost, I assume you've got FAS3240 single controller running in 7-Mode (Anyway, concepts are same and unified across all NetApp storage boxes. Only NetApp has pure unification from entry level box to high-end box in both 7-Mode and C-Mode)
Okay, let's go the business..
Rather than keeping each application's data in a separate aggregate, I would suggest you have a separate aggregate for root with RAID_DP using three disks (as you have said) and one single large aggregate for all your data needs. This will increase the performance of applications since large number of disks now contributes to applications' data.
You can have separate volumes (and LUNs inside those volumes) for each application and its data, logs, achieves etc. Additionally you may configure scheduled Snapshots independently for each volume that you create depending on your business requirement. Also, you may use thin provisioning and deduplication to achieve highest storage efficiency.
So, in summary, I may configure something like this...
Aggr0 - rg0 (1D+2P)
vol0 (i.e. Root Volume)
Aggr1 - rg0 (19D+2P), rg1(19D+2P)
Vol_VMWare (for VMware)
Vol_Exch (for MS Exchange)
Spare Disks = 3
"rg" stands for 'raid group'
Additionally, if you have the required licenses, you can use VSC (Virtual Storage Console) and Snapmanager for Exchange.
Here are couple of links for your further reading..
Hope above helps...
2013-06-11 07:46 PM
Hi,thank you for your response, information and sorry for my late reply.
Yes our FAS3240 has dual controllers running 8.1.2 7-Mode. 1 in primary site and 1 in secondary for DR so snapmirror is used to replicate current CIFS data.
I will look to go with this configuration as per above but before i do just 2 questions please.
Is it ok to mix VMWare and Exchange within the same aggregate spanned across all disks as they are 2 different styles of applications and how they write there data? Currently on our other storage provider we have a total of 54 spindles for both VMWare and Exchange but now with the new design as above we will be dropping to 38 so we are worried about performance.
Also do you know if its ok or best practice to run deduplication across the Exchange 2010 Vol?
2013-06-12 02:41 AM
Not seen the other part of the post here…
But in answer to your questions…
You NetApp environment should be sized for performance...don’t overly worry about the amount of spindles…if they are all part of one aggregate, you are getting the performance of pretty much all of those spindles to every volume you create.
If you add that to the fact that all writes actually go to Cache rather than disk, then your performance should be spot on… but all in the initial sizing to be fair…
As for your other question, absolutely no problem, assuming you are talking about de-dedupe of the volumes with the mailbox stores in?
If you are, not only a good idea, but almost a recommendation to do that, especially as there is no single instance in Exchange 2010 so you DB size is much bigger than previous Exchange versions, you won’t see any performance penalty for running de-dupe so if it was me and all things been equal, then use it…
2013-06-14 08:04 AM
To add to what Paul said, yes enabling deduplication on the Exchange database volumes is a very good thing and is inline with NetApp's best practices. Don't enable deduplication on the Exchange log volumes as that is unique and transitory data so the limited space savings isn't worth the processor cycles to run the deduplication scans.
Mixing workloads on a single large aggregate is one of the best approaches because you will 'average' out the workload by combining high IO workloads such as Exchange with lower workloads such as NFS mounts (assuming the servers aren't driving a high level of IO at the mounts).
Data is stripped across every data disk in the aggregate so IO should be evenly distributed and this is why putting all your data disks into one aggregate is a better choice.
With your configuration, I agree with SNETAPPUSER's configuration of 21 disk raid groups (19 data + 2 parity) and 2 raid groups in your data aggregate but I'm a little concerned as you say this:
"Yes our FAS3240 has dual controllers running 8.1.2 7-Mode. 1 in primary site and 1 in secondary for DR so snapmirror is used to replicate current CIFS data."
This makes me think you have one controller in location A and one controller in location B. Is this the case? And does EACH controller have 2 DS4243 disk shelves with SAS?
Or are these controllers in a MetroCluster configuration. If they are then the proposed configuration above might not work out as you need to have a plex0 and a plex1
2013-06-16 09:30 PM
Thank you for your response and i have also been reading Netapp Exchange 2010 best practice docs as per above which is very good. It makes mention to split the DB's & the logs into separate volumes / LUNS so ill be looking to do this.
I will most definitely now run dedup on the Exch DB volume to get the storage benefits that come with that and depending on my total storage requirements that come back go with a configuration similar to this.
Disks total = 48 - 600GB SAS
Parity = 6
Spare = 4
Disks Useable = 38
vol0 (i.e. Root Volume)
Vol_VMWare (for VMware)
Vol_Exch_DB (for MS Exchange DB's)
Vol_Exch_Logs (for MS Exchange Logs)
If there is anything I've missed i'd love to hear from you.
2013-06-19 12:01 AM
I was hoping 2 questions clarified as im struggling to find this in the documentation.
1) To clear the exchange logs can nightly snaps on the DB volume complete this task or is that a job for Snap Manager for Exchange?
2) We will have 1 volume for DB's and 1 for logs so is it best practice to thin provisioned these?
2013-06-19 05:17 AM
1) No, you need to have an agent of some sort that interacts with Exchange to quiesce the application to create an application consistent (i.e. recoverable) backup. It is this interaction that can tell Exchange to truncate the logs after the backup has been performed. Snap Manager for Exchange (SME) performs these functions and every backup software package/appliance that supports Exchange does also.
2) I'm not sure about 'best practice' as you will hear arguments for both sides. For me this is definately a YMMV (Your Milage May Vary). If you are over provisioned on your LUNs/Volumes for Exchange and don't expect ALL of them to reach a high level of utilization, then thin provisioning is a valid option as that will give you a pool of free capacity on the aggregate(s) that one or more FlexVols can use when needed. Thin provisioning is writing an IOU for space that you a) hope never gets cashed or b) by the time it does get cashed, you have pruchased additional disk to support it. For example, in our environment our Exchange LUNs are provisioned at 750 GB but use about 250 GB. So the LUNs are thin provisioned in the FlexVols and the FlexVols are thin provisioned on the aggregate and I can have about 5 Exchange DB Luns reach near 100% utilization of space before I start to get really nervous about free space in the aggregate.
Hope this helps.